Yes, well put. I agree that the ability of TR to select appropriate workouts for me is nothing short of amazing, and it is this feature that has kept me using TR for years. Thank you for reminding me of what is most important.
When I posted, I also did not realize that I could turn off the predicted FTP, but now that I realize this, I am happier.
Similar issues here. My frustration is that these are not user errors or settings to be toggled on/off and people aren’t making this up. Your ftp does not change 10% in 5-10 days based on one workout. Thats not how real life works so why is it portrayed that way.
Regardless if it’s user error, account setting preferences, or just discouraging predictions, I hope the TR tech people that actually work on the development are hearing these things and addressing them. They have what I have always felt to be an amazing product, but I feel like these are usually dismissed issues in the forum by people who have not had the same issues.
I appreciated that Caro commented on my similar post to explain, but I also would like some encouragement to stick it out as they are working on fixing some of the anomalies.
My other self criticism is that I should have just emailed support instead of the forum as I don’t think anyone on here can fix anything, just provide workarounds, unless they are TR programmers reading it
Quick context on my issue, like yours was 1. Weird predicted fluctuations in the FTP but also 5 straight days off after an A race that was only 60 min long. I don’t know anyone that would prescribe me 5 days off after an A race, no matter if it was an A B or C event. Unless it was the end of the season and I had planned to take time off. However, this is not the case, since I have B races in the weeks following and “AI” could see this and had no regard for it. Just some strange stuff.
Hey @Niterider4 
Do I have permission to share screenshots of your Calendar? There is a very good reason as to why your FTP has gone down, but I would need to share screenshots to explain.
If not you can DM me or email support@trainerroad.com and we can help you directly and explain things further.
Absolutely, you have my permission. I’m happy to learn more.
What do you do with your other rides?
For this workout plan, for which I am at week 5 of 13, I do not have any other rides. I am only riding on TR, 4x a week.
Same here, initial prediction 320W, next one in 3 days 298W… It will definitely go down after tomorrow‘s workout too. I haven’t missed hard workout and struggle to understand what on earth I would need to do to meet the 320W prediction. Every harder workout I tried adding to calendar resulted as FTP decrease. I am not saying 298W is wrong, 320W is.
It’s in Beta and maybe things are getting dialed in behind the scenes? I’d take the new (and if I’m gathering from your comment more accurate) number and see what happens. The big thing is are the workouts effective and progressing your training?
I realize why my data isn’t accurate but my experience has been the opposite. I just wish they kept the previous FTP auto detection as an option for those of us that the new AI FTP doesn’t work for. I got an 8 watt predicted increase from skipping two workouts over the weekend.
@Jonathan mentioned in the last podcast that some people are better off training at a higher FTP and some lower. I wonder with if that is what is in play here?
It would be an interesting concept to dig into for a podcast. Is this tied to training history, volume, power profile?
If the ai has no concept of FTP then why would adjusting your FTP make the training better/worse?
It just seems very strange to me that an AI engine would predict a change in FTP of 16.5% over 20 days and 16 workouts. Every bike rider on the planet knows that the likelihood of that kind of a change is pretty much zero, so how is it that TR’s AI engine is spitting out projections like this as a matter of course?
I like the idea of an FTP projection, but it seems that the projections TR is giving at the moment bear no semblance to reality.
I always found that the automatic FTP detection was pretty realistic, so it is strange that the AI predicted FTP is so wonky.
I also took note of that comment and wished he expanded on it. I wonder if that has to do with folks who have more natural anaerobic contributions vs well developed sustained LT2 bias.
I was also wondering about my own situation where my AIFTP is much higher than my actual FTP. I suppose it’s possible that the ML saw something in my data that thinks I respond better to training more based on critical power or even higher.
I hope we get more clarity on that comment at some point.
Agree. I’d love for @Jonathan and @Nate_Pearson to have another podcast a couple months in. What have they learned? Where is it going? I love it but also my prediction isn’t very variable. When it goes down it’s because a workout didn’t go well or I blow off a workout for a zwift race.
Because the workout library is organized around FTP which turns AIFTP to something more like a training intensity factor than anything else. It’s basically an intensity/TiZ slider in my PoV. With an 1h constraint, which is what many people seem to have, it pushes towards intensity.
Edit: I raised one concern some months ago and I’m still unsure how things stand. Does the algo optimize for AIFTP (the intensity factor)? If so, I’d be worried about neglecting parts of the power curve. I can’t tell from experience because I’m using a manual entry.
Based on what I’m seeing with your recent training, it looks like things are going really well for you… 
You’ve had some ups and downs with your training over the past few months, which naturally brings fluctuations to your fitness. You saw a couple of drops in FTP between January and March, but now things are on their way back up.
All of your post-workout survey responses appear to be spot on to what we’d expect, so your workouts look like they’re dialed in.
I’ll be curious to see how Sunday’s workout goes since it’s your second threshold workout since your last FTP bump. Hard starts are tough, but you made it through last week. 
I’ve taken a look at what we think your FTP should be if you had an FTP Detection today, and it’s exactly where I’d expect it to be.
Sunday will tell us if your current FTP is set correctly, so follow up here if you could afterward.
Let me know if this helps and good luck! 
This is very helpful. Thank you for looking at it.
In January and February I was backcountry skiing, so I was staying in shape. I just wasn’t on the bike.
So I completed the prescribed workout today (Hard Starts) and, once again, my AI predicted FTP went down, this time by 0.9%.
I am following my TR prescribed program to the letter, on a super high end erg bike, with no other workouts, and every time I complete a workout, TR tells me I am less fit than I was before.
It seems to me that, as other people have reported, TR starts out with an AI predicted FTP that is stupidly, unrealistically high and then, with each successive workout, TR lowers the AI predicted FTP until it eventually arrives at a number which bears no resemblance to the original AI predicted FTP number.
For me, this behaviour does not resemble AI, nor is it what I am looking for. From my perspective, this is a step backwards for TR.
A prediction is the future. Is the prediction higher or lower than your current value?
Earlier you were showing a 16.5% increase are you still showing 5% plus?
The prediction is now 0.9% lower than my current value, which does not make sense to me.
On Jan. 22, I was at 227, which seemed right to me.
On Feb. 22, I was at 209, which reflects that I was not riding that month. However, I was backcountry skiing every day, which got logged on TR but was not reflected in my FTP on Feb. 22.
On Mar. 22, I was back to 227, but within a day or two my predicted FTP was 241, which was ridiculous, as I have never been higher than 235. Since then, my predicted FTP has been floating downwards with every workout, even though I have been following the plan to the letter. My predicted FTP is now 225.
It would seem to me that on Mar. 22, TR should have predicted my FTP at something like 230, and then just stuck near that number while I do all my prescribed workouts. Instead, my predicted FTP was immediately crazy high and is now too low. The predicted FTP is pretty much useless to me, because it just seems random.
